Types of Machine Vision Systems and What They Can Do

What are the main types of vision systems?

While machine vision systems offer a wide range of features and options, there are three main categories to consider.

Line-scan cameras are used in continuous inspection applications such as web manufacturing. They take a wide but very thin image, typically as material moves past the scan area, and use software to reconstruct the image line by line. They are significantly faster in these applications than standard 2D cameras. Examples include inspecting fabric, paper, and other soft goods.

Most machine vision uses 2D cameras, also called area scan systems. These can be simple, single-purpose sensors or more full-featured systems. Sensors are lower cost, smaller, easier to deploy, and often more rugged than cameras with a wider range of features. Generally, full-featured systems are more configurable, perform more complicated tasks, and can take larger and higher resolution images. The decision between sensors and more complex machine vision systems comes down to the task being performed, the format of the data output, cost, and ease of use. 

3D vision systems add depth to their images, sometimes using lasers to measure distances and compute depth. The addition of depth can increase complexity and cost to some degree, but that additional information is critical to some applications. For example, 3D vision is essential in guiding a robot arm to precisely reach out to grab an object in the correct orientation, no matter where it is in space. 3D vision is also used to automate difficult cutting and welding processes. 

What does machine vision do?

Machine vision providers sometimes classify what their systems can do with the acronym GIGI, for Guidance, Identification, Gauging, and Inspection. This guide breaks down the types of applications into slightly more detail:

• Defect detection
• Object detection and counting
• Measuring/gauging
• Locating/guiding/positioning
• Barcode reading
• OCR/OCV

Defect detection

Defects can occur in any part of a manufacturing process, from problems with the quality of raw materials or parts through final inspection. Inspections to catch defects have traditionally been done manually by trained workers. 

Machine vision is a significant improvement over human inspection: it operates at production line speeds, doesn’t get tired, can detect even small and unexpected defects, and stores information for continuous operational improvement.

Examples:

Object detection and counting

Determining the presence or absence and counting objects is a widely used function in inventory management, on production lines, and before releasing or accepting shipments. Both manual inspection and mechanical counting are slow and prone to error, compared to high-speed and consistently accurate machine vision systems.  

Examples:

Measurement & Gauging

Precise manufacturing requires accurate measurement of distances, areas, diameters, and more. Manual measurement with gauges, calipers, or inspection jigs is slow and introduces inconsistencies.

Machine vision consistently and accurately measures down to the micron level, while parts are on the line, at high speeds. Each image and its associated data can be stored in case of warranty or compliance issues.  

Examples:

Locating, guiding, and positioning parts

Functions such as assembly, pick-and-place, and inspection depend on machine vision’s ability to locate a part, whether on a conveyor or in a bin. Machine vision guides robots in picking and placing parts and ensures accurate micron-level positioning in precision assembly. 

If a part is out of place, machine vision measures the difference between desired and actual location and orientation and communicates that to a robot or programmable logic controller (PLC) to realign the part.

Examples: 

Barcode reading

Barcodes are used to track and identify raw materials and finished goods throughout the supply chain. They include 1D and 2D codes, such as UPC codes on retail products and data matrix codes on packages. 

Barcode readers process information at high speeds, and correctly read codes that are partially torn, obscured, smeared, or distorted. Image-based barcode scanners collect images of the barcodes they read, allowing for analysis of no-reads or misreads to diagnose problems, such as a clogged print head, damaged codes, or inadequate lighting. Multiple camera systems can scan from multiple sides of a part or package at once to increase read rates when not all codes are oriented the same way.

Examples: 

Reading text: OCR and OCV

Barcodes are everywhere in modern industry, but human-readable printed text is still essential, particularly for retail, manufacturing, pharmaceutical, and food and beverage supply chains, to identify sell-by dates, lot numbers, and other important information. For this text to be useful in modern high-speed processing, it must be reliably readable by machines. Machine vision systems can read these codes in milliseconds with 99.99% accuracy. 

Optical character recognition (OCR) and optical character verification (OCV) both identify and interpret text in images, with one key difference. OCR reads text to trigger another action, while OCV is used to verify the quality of text against a known standard.

Examples: